E21 - Macroeconomics: Consumption; Saving; WealthReturn

This research is related to the assessment of alternative unemployment rate predictions for the Romanian economy, the forecasts being provided by three anonymous forecasters: F1, F2 and F3. F3 provided the most accurate forecasts for the horizon 2001-2014, while F2 predictions are the less accurate according to U1 Theil's statistic and according to a new method that has not been used before in literature in this context. The multi-criteria ranking was applied to make a hierarchy of the forecasters regarding the accuracy and five important accuracy measures were taken into account at the same time: mean errors, mean squared error, root mean squared error, U1 and U2 statistics of Theil. The combined forecasts of forecasters' predictions are the best strategy to improve the forecasts accuracy. The filtered and smoothed original predictions based on Hodrick-Prescott filter, respectively Holt-Winters technique, are a good strategy of improving the accuracy only for F2 expectations. The assessment and improvement of forecasts accuracy have an important contribution in growing the quality of decision-making process.

This paper introduces a model for stress testing of probability of default of individuals. The model rests on assumption that the individual defaults if his savings fall below zero. The probability of default is then described as a function of several macroeconomic indicators, such as wages, unemployment and interest rates. Stress testing is carried out by applying exogenous stress scenarios for development of these indicators. The model implies that sensitivity of probability of default to the stress is mainly driven by installment to income ratio and for mortgages also by loan maturity. Hence installment to income ratio is suggested as the appropriate tool to manage credit risk of retail portfolios.